Many advanced semiconductor electronic and opto-electronic devices are being developed based on the InP family of semiconductors. An example is a fast transistor using InP channels that take advantage of the high saturated electron velocity in InP. In this case, In.sub.0.52 Al.sub.0.48 As is grown over an InP substrate or InP epi-layer. The band bending at the InAlAs/InP interface leads to two-dimensional confinement of electrons on the InP side of the interface and of holes on the InAlAs side. Another example is a thin quantum well of InAs grown on an InP barrier and then covered with another InP barrier. The bandgap in the quantum well corresponds to about 1.3 .mu.m of optical energy, a region of great interest for optical fiber communications. In both examples, the interfacial characteristics are crucial because the interesting effects are occurring within a few nanometers of the nominal interface.
Such structures can be grown by a variety of techniques, molecular beam epitaxy (MBE), organo-metallic chemical vapor deposition (OMCVD), or organo-metallic molecular beam epitaxy (OMMBE), which is a combination of the first two and is sometimes referred to as chemical beam epitaxy (CBE). The conventional OMMBE method will be described with reference to an InAs/InP interface. An InP epi-layer is deposited by exposing an InP substrate to a medium pressure of trimethyl-indium (TMI) by the evaporation of TMI (a solid at room temperature) and to a medium pressure of molecular phosphorus (P.sub.2) obtained by cracking phosphine in a cracker cell. The TMI cracks on the hot substrate surface into In which then combines with the arriving P.sub.2 to form the desired InP. At the interface, the beam of phosphorus is interrupted and is replaced by a beam of molecular arsenic (As.sub.2) from another cracker cell cracking arsine to similarly form the desired InAs. If the second layer is to be composed of InAlAs, tri-isobutyl-aluminum (TIBAl) is supplied by the evaporation of TIBAl (a liquid at room temperature) in a fixed proportion to the TMI. Because the absence of the group-V flux is believed to degrade the surface quality, it is common practice to interrupt the TMI for about two seconds during the switching from P.sub.2 to As.sub.2 and before supplying the TIBAl in order to assure that the group-V element is not a combination of P and As.
Our experience, however, has shown that the InAs/InP interface is less abrupt than would be expected from the above sequence. We attribute the major part of the poor interface to an exchange between the already grown phosphide with the after supplied As. Indeed, we have grown a layer of InAs that was between 5 and 10 monolayers thick by simply exposing the InP to As.sub.2 for eight seconds in the absence of TMI. The thickness though was spatially varying. Such interfacial roughness undesirably scatters electrons, thus limiting device performance. It is possible that such As-P exchange continues after the fabrication of the interface, thereby bringing into question the long-term reliability of a device utilizing such an interface. Obviously, the poor interface abruptness and questionable reliability are not satisfactory.
A similar problem has arisen in GaAs/GaInP heterostructures, which are important for modulation-doped field-effect transistors, heterojunction bipolar transistors, lasers, and solar cells. The alloy Ga.sub.0.515 In.sub.0.485 P is lattice matched to GaAs. A two-dimensional electron gas forms on the GaAs side of the heterojunction, as disclosed by Razeghi et al. in "Extremely high electron mobility in a GaAs-Ga.sub.x In.sub.1-x P heterostructure grown by metalorganic chemical vapor deposition", Applied Physics Letters, volume 55, 1989, pp. 457-459. In a superlattice configuration, GaAs forms the quantum well. Such GaAs/GaInP heterostructures can be grown by OMMBE, by gas source MBE (GSMBE), or by OMCVD. The latter method is often preferred because of its simplicity, economy, and high throughput. However, changing over the group-V element is known to present a problem. The P cannot be shut off in a cessation of growth during which the chamber is flushed because its high vapor pressure would cause it to evaporate from the already deposited GaInP. Bhat et al. have described the deleterious reaction of arsine and phosphine with underlying III-V materials, especially GaAs and InP, in "A novel technique for the preservation of gratings in InP and InGaAsP and for the simultaneous preservation of InP, InGaAs, and InGaAsP in OMCVD," Journal of Crystal Growth, volume 107, 1991, pp. 871-877. That article disclosed that neither InP nor InGaAs is stable in arsine or phosphine respectively. Thus, when GaAs is grown on GaInP, the conventional technique of immediately substituting arsine for phosphine is likely to produce an intermediate layer of GaInAsP of graded composition and having an uneven surface. Such a layer can have a lower bandgap than GaAs and so will capture free charge carriers, principally heavy holes. Therefore, recombination occurs close to the interface where the material's wavelength is longer than desired. Furthermore, the efficiency is likely to be reduced because of nonradiative centers arising from the interfacial roughness. Also, the graded composition smears the interface, a critical problem when the GaAs layer is thin so as to serve as a quantum-well layer. That reference also suggested that As.sub.4 rather than As.sub.2 should be used to preserve InP during its heat up before growth.